To investigate the association between neonatal phototherapy use and childhood cancer. This retrospective cohort study included 499?621 children born at ?35 weeks’ gestation from 1995 to 2011 in Kaiser Permanente Northern California hospitals, who survived to hospital discharge and were followed ?60 days. We obtained data on home and inpatient phototherapy, covariates, and cancer incidence from electronic records. We used propensity-adjusted Cox and Poisson models to control for confounding and unequal follow-up times. There were 60 children with a diagnosis of cancer among 39?403 exposed to phototherapy (25 per 100?000 person-years), compared with 651 of 460?218 unexposed children (18 per 100?000 person-years; incidence rate ratio [IRR] 1.4; P = .01). Phototherapy was associated with increased rates of any leukemia (IRR 2.1; P = .0007), nonlymphocytic leukemia (IRR 4.0; P = .0004), and liver cancer (IRR 5.2; P = .04). With adjustment for a propensity score that incorporated bilirubin levels, chromosomal disorders, congenital anomalies, and other covariates, associations were no longer statistically significant: Adjusted hazard ratios (95% confidence intervals) were 1.0 (0.7-1.6) for any cancer, 1.6 (0.8-3.5) for any leukemia, 1.9 (0.6-6.9) for nonlymphocytic leukemia, and 1.4 (0.2-12) for liver cancer. Upper limits of 95% confidence intervals for adjusted 10-year excess risk were generally <0.1% but reached 4.4% for children with Down syndrome. Although phototherapy use was associated with increased cancer rates (particularly nonlymphocytic leukemia), control for confounding variables eliminated or attenuated the associations. Nonetheless, the possibility of even partial causality suggests that avoiding unnecessary phototherapy may be prudent.